Research On Swan-inspired Visual Inertial Odometry And Flight Control Of The UAV

In recent years,the autonomous positioning of drones and the three-dimensional perception of the environment have been becoming important research hotspots in the field of robots.The autonomous positioning and sensing of drones in the GPS-denied environment has become a key technology.At present,the positioning of UAVs all over the world is achieved by GPS or ultrasonic sensors.It is impossible to achieve positioning when the GPS signal is not well.In order to solve this problem,many research institutes have proposed cameras and inertial measurement units(IMU)on drones,realizing visual inertial positioning of drones by these two sensors.However,the camera and IMU are fixed to the body of the drone,which limits the movement of the camera.Meanwhile the view of the drone have to change with the movement of its drone body.When the drone performs various tasks such as hovering,its field of view remains fixed and the environment outside the field of view is not detected,which greatly limits the visual sensing ability of the drone.In view of the above problems,this paper develops a swaninpired UAV visual inertial odometry and flight control system.The system mimics the visual positioning of the swan.The three degrees-of-freedom(DOFs)manipulator mounted under the drone is considered as the "neck".The camera and IMU fixed at end of the three DOFs manipulator are considered as the "visual inertia sensing system" of the drone.Firstly,analyzing the time stamps of different visual inertial sensors,and then selecting the monocular camera and IMU with good time-synchronization performance.Secondly,a three DOFs manipulator with high-precision position feedback is designed and the forward kinematics relationship of the manipulator are modeled.What’s more,the odometry positioning information is transmitted to the UAV’s body through the forward kinematic relationship of the mechanical arm.Finally,separating the camera from the UAV body,the camera can move by controlling the movement of the manipulator,and then realizing the multi-degree-of-freedom flexible pose estimation of the drone.In order to verify the feasibility of the method in this paper,a flight experimental platform is built and different flight experiments are conducted.By comparing the visual inertial odometry with the information of the motion capture system,the visual inertial positioning algorithm used in this paper achieves the millimeter-level positioning accuracy,which can fully meet the position feedback requirements of the drone.By comparing the positioning information of the UAV body with the motion capture system,the swan visual inertial positioning system built in this paper has high positioning accuracy.The swan-inpired visual inertial positioning and control method proposed in this paper greatly improves the flexibility of the UAV visual positioning and enhances the perception ability of the UAV.It has very important research significance.